Minimizing TCO in Information Systems

Starting in 1996, there has been much talk about the total cost of ownership (“TCO”) of personal computers. For example, in that year the Gartner Group estimated that the TCO for a networked PC is approximately $11,900 per node per year for hardware, software, support and administrative services, and end-user operations. This high TCO has permitted Oracle and Sun to tout the network computer, which they claim will provide meaningful decreases in total cost of ownership. However, the high TCO of PCs is primarily due to bad management rather than inherent deficiencies in the PC. Companies can reduce the TCO of PCs by following these practices, which I call “Mitchell’s 41 Steps to Reduce TCO”:

Standardize on one microprocessor. In most cases this should be Intel. Microprocessors such as PowerPC and MIPS offer little or no advantages over Intel microprocessors, and have much higher risk profiles.
When you purchase servers, purchase more than you need and make certain you can add additional processors, memory and hard disk to each of your servers.
The price of a server should be the last factor you consider. Much more important is performance, expandability, reliability, fault tolerance, ease of setup, and ease of management. In most cases, a Dell PowerEdge or Compaq Proliant server should be chosen.
Standardize on one server/network operating system. In most cases this should be Microsoft XP Professional for desktops and notebooks. For servers, depending on the circumstances, it could be sense to use Windows Server or Linux.
Unless you have specificized needs, I would only consider purchasing Desktops, notebooks and servers from Dell, IBM or Hewlett-Package.
Choose one product line from that manufacturer that provides system management features and consistent features and peripherals. Dell, for example, has two desktop product lines: the Dimension and OptiPlex. The Dimension computers use state-of-the-art peripherals and are meant for home and business users that will not be networking their computers. The OptiPlex computers offer a more stable platform and are designed for network use. Businesses intending to network their PCs will therefore want to purchase an OptiPlex over a Dimension.
Standardize on one (or at most two or three) desktop hardware configurations. Your life will be much easier if, for example, everyone is using the same video card and monitor.
Choose a high end, leading edge hardware platform as your desktop standard. Purchase a workstation rather than a PC. Doing so means that you can use your computers for 12 to 18 months longer than if you chose a medium end platform. Remember that the initial cost of purchasing a personal computer or workstation is typically less than 15 percent of the TCO. A high end NT workstation costs approximately $7,000 to $8,000.
Remember that purchasing leading edge technology does not mean purchasing bleeding edge technology.
Install every peripheral you might need and more than enough memory and hard disk. Because of the labor and configuration costs, adding a second hard disk 1 year later will cost you considerably more than doing when you purchase the computer, even after you account for the decrease in storage costs.
Purchasing components and then assembling your own PC is not a good use of your time. The leading manufacturers add significant value by testing the interactions of various components, the cost of which they spread over millions or even tens of millions of computers. It is foolish to do this yourself.
Standardize on the printers you use.
Standardize on one desktop operating system. Many large companies have DOS, Windows 3.1, Windows 95/98, Windows NT, OS/2, Macintosh OS and Unix on their various desktop computers. Rarely is there any business advantage in running so many operating systems. In most case Microsoft Windows 2000 Professional should be chosen as the desktop standard. (Installing Windows 95/98, only to upgrade a few years later, would be foolish.)
Run the same version of the operating system on all of your desktop PCs; everyone should upgrade at the same time.
Standardize on the software packages you use. There is rarely any good reason, for example, for an organization to use more than one word processor or spreadsheet.
In many cases, choosing a suite of packages makes the most sense. First, you save a lot of money by purchasing the suite rather than the individual packages separately. Second, suites usually offer similar commands and functionality among their component packages, reducing your learning time. For a desktop productivity suite, Microsoft Office should be chosen. For a server suite, in most cases Microsoft BackOffice should be chosen.
Choose wisely which software packages become your corporate standard. In addition to looking at feature sets, you should consider whether the fundamental design and architecture of the package will permit it to grow in terms of functionality. The financial stability of the software publisher is important, as is the level of R&D resources being spent to enhance the package. The infrastructure = technical support, books, periodicals, training courses, consultants and add-on packages = supporting such package should be carefully considered. In addition, pay particular attention to how a software package will integrate with your other software packages. In the 21st century, integration will become the most important feature of software packages.
When you purchase a computer, install all at once all of the software packages you will need. Better yet, let the hardware manufacturer or your computer reseller install the software for you.
Since the Windows platform is so difficult to manage, it makes sense to have a test PC on which you install all software being considered. Only after a package works properly on your test system should you consider installing it on your production PCs.
Standardize on versions of the software packages you use; everyone should upgrade at the same time. In many cases, different versions of the same software package will use different file formats, which will make use of files by several people difficult.
For mature software packages, such as word processors and spreadsheets, do not upgrade every time a new version comes out; rather, upgrade every other version. Obviously this should not be done for immature software packages where it is essential to get the new features offered by each new version.
When you upgrade a mature software package, do so at the later of (i) 6 months after the product is introduced or (ii) the release of at least one maintenance release of that version. If you wait at least 6 months, it is likely that the infrastructure for that version will be well established. By waiting at least one maintenance release, you will reduce the likelihood that any non-trivial bugs exist in that version.
One person should be in charge of installing and upgrading software. It makes no sense, for example, for 100 users to download the latest version of Firefox from the Internet.
Standardize on the configuration options you select when installing software on servers, desktops and notebooks.
If you will be installing more than 200 copies of a software package, use a system management program such as Microsoft SMS Server.
Develop a well thought out and consistent subdirectory structure for your server. Document this structure and communicate it to your users. Users should be consistent about where certain files are saved.
Standardize on the subdirectory structure you use on local hard drives. If you are installing Microsoft Office on 100 desktop PCs, for example, Office should be installed in the same subdirectory on all machines.
Use long file names; short file names (such as the DOS 8.3 file name structure) and medium file names (used by the Macintosh OS, which does not allow file names longer than 32 characters) should be avoided.
Develop standards for naming of files, and require that everyone follow them.
Users should not save data files to their local hard disk, since users rarely backup their local disks. Rather, all data files should be saved to the server, and a regular backup should be made of the server.
In calculating TCO, include all costs, including installation, supplies, maintenance and electricity. If, for example, you are purchasing an Iomega Jaz drive, ask not for the cost of the drive, but the total cost of the drive and having it installed and tested. Many network laser printers, for example, will cost more in supplies than their initial cost. Energy efficient hardware can save many dollars in electricity.
Set up templates that permit your employees to start with a standard file rather than a blank file. This allows them to be productive right away.
Be skeptical about participating in beta tests of new software. Rarely will the benefits you receive justify the hassles.
Provide good training for new employees.
Provide good training every time you upgrade versions of software packages, even if it is only a 2 to 4 hour course.
Consider having on-line technical resources available for your users and developers through networked CD-ROMs attached to your server. Microsoft TechNet, for example, is inexpensive and can answer almost any question one might have concerning Microsoft products.
Maintain a comprehensive library of documentation and manuals for the software packages you use. The quality of computer books has increased dramatically since the PC was introduced in 1981.
Write an internal computer manual which provides technical information not available from other sources. Keep it up-to-date.
Provide (either internally or externally) good technical support for your users. If a user waits 20 minutes on the telephone when calling technical support, and then is given the runaround, that will cost your company money.
Each category of your IT/S expenditures should be examined to determine if it should be outsourced.
Last but not least, always think in terms of TCO rather than initial cost. Remember that for every dollar you initially spend on a computer, you will spend several dollars over the life cycle of the computer.

If you follow these strategies, the total cost of ownership of PCs can be close to the cost of network computers, and at the same time you will also have all of the advantages of PCs.

The Gartner Group has published its Best Practices List to reduce TCO:

Asset Inventory. Non-automated inventories are 20% inaccurate on average, resulting in mis-statements within accounting data, increased losses due to theft, mis-allocation of assets, and incorrect tax reporting on IT assets. Using asset inventory, these problems can be reduced or eliminated. As well, up to 50% of each help desk call is spent on understanding the hardware and software configuration. Call time can be reduced by providing access to the inventory data to the help desk support personnel.

Virus Protection and Elimination. With the proliferation of data sharing via e-mail and the Web, the chances of virus infection has increased dramatically. New viruses infect systems, documents and even cells in spreadsheets. Virus detection and elimination helps to maintain user productivity and eliminate costly administration labor in repairing virus infections. Viruses can take up to 14 person hours to eliminate from a network and cause 4 hours of downtime for each infected machine. An organization with 1000 computers is expected to encounter at least seven infections over a 12 month period. A proper virus protection plan and tools eliminates the risk of infections and provides a means to more quickly recover should an infection occur.

Centralized Network Management and Desktop. By implementing network management products and centralizing the management of performance information and alerting, administrators can better set policies, pro-actively avoid issues and respond quickly should issues occur. This includes DMI, WFM compliant desktops and management systems.

Centralized Help Desk, Support, Knowledge Base and Remote Control. By consolidating calls to a central location economies of scale in the call center can be achieved. Problems are better centrally tracked with common problems being identified and pro-actively eliminated. Combined with an inventory system that can display configurations of the system having issues and a desktop remote control capability, the help desk can be dramatically improved, reducing the duration of calls by 50-70%, and reducing call frequency.

Beware, however, that help desk support needs to continually monitor service levels to assure performance. In most organizations, help desks do not provide adequate knowledge and response times to meet user expectations. Users then turn to peers for support.

Scalable Architecture. Business is continually changing and competition continually increases. As a result, the organization you manage today is not the same one you managed a year ago. It is vital, in order to reduce the cost of keeping up with change, that you develop an architecture that provides for scalability and flexibility to meet the ever changing business needs. Re-engineering of the infrastructure on a continuous basis is costly. Select an architecture that is scalable to reduce overall change costs.

Component Application Software. Similar to a scalable architecture, it is important to create a scalable application development environment, one which can quickly deliver initial products, build to meet growing business needs, and flexible enough to meet changing business requirements.

Fault Tolerant Systems and Network. If network resources were available 96% of the time, better up-time than most networks experience, a 300 user network would still be faced with losses of $840,000 each year in lost productivity, the inability for your users to accomplish their work, and lost revenue, the bottom line revenue impact of not being able to perform mission critical transactions. A fault tolerant network that implements fault tolerant storage (RAID), servers, and network components can increase uptime and eliminate costly resources.

Policy-based Management. Implementing policy-based management limits the access of users to specific applications, data, and system control needed for their jobs. By limiting the users to just the right amount of computing power, management and support costs are reduced, while productivity and scalability is maintained. Policies eliminates much of the end user IS costs by providing a system that does not allow users to access application for which they are not trained. With policies, much of the end user IS costs can be eliminated, providing for a system that does not allow applications to be run on which a user has not been trained, as well as eliminating system changes.

Standardize Your Desktops and Network. A standardized desktop platform and network architecture will allow for economies of scale in purchasing, support, and management. Purchasing can be performed in volume to get better pricing. Components can be quickly replaced because they are swappable and spares can easily be made available. Support and management are reduced because each system and network is identical, reducing training and re-education expenses, and allowing common issues to be identified and engineered out of the organization in a unified fashion.

Backup Data and Plan for Disasters. Most organizations backup their server data, but do not backup client information, and rarely are prepared with continued operation or recovery plans in the event a disaster strikes. Many keep the backup tapes next to the server being backed up … in a site disaster the server and the tapes are both destroyed. A good disaster recovery plan includes tested policies for data protection, emergency team formation and alternate site preparation and procedures. Specialized consultants and services are available to assist in developing and implementing these plans.

Better Planning of Projects. A better plan and up-front analysis avoids costly mistakes later. A team oriented planning methodology provides structure to the planning process. A plan involves multiple members from representative organizations with defined roles, vision, scope and most importantly a method for identifying and mitigating risks. TCO management is one planning method that lets you plan IT finances proactively by analyzing issues, identifying trouble spots, and simulating plans prior to improvements.

Centralized and Streamlined Purchasing Procedures. The capital expenditures of computer hardware and software can be greatly improved by centralizing the purchasing of all assets, and obtaining maximum cost savings through volume license and hardware purchasing agreements. However, centralized purchasing can cause issues and cause the user to circumvent the system if too many restrictive policies are introduced coincident with the project. It is vital to balance the cost savings against the natural tendency to restrict control and flexibility.

Measured Service Levels. By reducing costs it is often the case that organizations do so at the expense of service to the business units and users they are supporting. As part of any cost reducing projects that you implement, service levels should be monitored and minimum required service levels should be established. A cost reducing program that drops service levels beyond the minimums is one that should be re-considered establishing a written service level agreement and monitoring the quality of the delivered services is essential.

Capacity Planning and Load Balancing. Adequate scaling of resources and allocation of computing power to meet user and customer needs is essential. Too often, resources are purchased and managed for capacities that may be too low, or too high. This causes mis-appropriation of hardware and software expenses, and worse, downtime or over-management of resources that are heavily taxed and are over-capacity. By monitoring the capacity and performance requirements of the business and users, resources can be allocated and maximized for needs.

Change Management and Control. Vendors and users are continually putting pressure on the IS organization for change. This demand is only going to increase over time. It is essential that an IS organization realize that change demands are continuous and put in place a system to control and manage change in an orderly and planned fashion, avoiding users from skirting the system to implement the systems and applications they require because IS can’t keep up. It is key to embrace, manage and control change rather than fight it.

Upfront Investments in the Right Systems. An organization that invests more in technology up-front can eliminate costly upgrades later on. Buying an extra 8 megabytes of memory per machine can cost $50 per system today. But an upgrade later will require not only the purchase of the memory, but the labor associated with purchasing and installing the memory, an estimated expense of $250 to $700 per computer. Similar expenses can occur by under-specifying hard drives, monitors, multi-media, network resources, printers, and other similar assets. Spending up front for the right systems can eliminate costly labor expenses later.

Cycle Technology to Match the Useful Lifecycle. Typical organizations attempt to hold onto computer assets for as long as possible. The extra expense of troubleshooting, repairing, upgrading and maintaining older computer assets for many organizations far outweighs the benefit of a 4 to 5 year lifecycle. It is recommended that desktop computers should be replaced after three years and mobile computers be replaced every 18 months. This extra capital expense will have returns in lower administration and support expenses, and should yield gains in productivity as the new PCs are able to run newer operating systems and applications. Leasing is also a viable option to assure that initial capital outlay is low, and technology is cycled on a continuous basis to match the lifecycle. Many leasing companies offer great rates, track assets, and include maintenance contracts for minimal fees.

Software Usage Monitor and Licensing. How many of the applications are really being used by end users? Many IS professionals do not know who is using the installed software and how many licenses are purchased that didn’t need to be. By monitoring usage, consolidating the software, and negotiating usage based licenses, you can optimally purchase software licenses and reduce software expenses.

TCO Lifecycle Management. By measuring TCO on a regular basis, tracking the cost issues, progress of cost reductions, and verifying returns from IT investments you can optimize budgets and spending.

End User Training. Training is one of the most important ways to increase productivity and reduce both direct and indirect IS costs, particularly helpdesk and peer/self support. You may wish to consider requiring formal training prior to new application or platform rollouts, and implementing a just in time training system to provide the skills and help information when a user needs it on operating systems, network, and applications.

IS Training and Certification. Support personnel must be trained on the operating systems, computers, network and applications they are supporting to adequately provide value added services. You should review certifications and implement incentive programs to increase the number of certified team members. When outsourcing of application platform and operating system rollouts, skills that are gained during these experiences are not transferred to the help desk and administration staff. It is vital to re-tool skills as the network changes and to be sure that outsourcing is performed in a cooperative way to assure knowledge transfer.

Motivated IS Professionals. A service organization (such as IS and help desk support) requires motivated individuals who are dedicated to meeting customer expectations and delivering superior service. An IS professional who is not happy with his/her job and is not motivated to delivering superior service will not be able to perform up to user and business unit expectations. This is costly in both terms of poor IS staff productivity, lost user productivity, and in the quality of delivered services.

Read James’ essay on spam.

List of other essays written by James Mitchell | Copyright notice

Cite as “Minimizing Total Cost of Ownership in Information Systems ” by James Mitchell. September 6, 2002, version 1.4.
www.jmitchell.me/essays/minimizing-tco-information-systems.